Built an AI-powered Park Finder Copilot using Retrieval-Augmented Generation (RAG) to answer natural language queries like “quiet parks near me with trails”, combining geo-location data with contextual recommendations.

Implemented semantic search using embeddings + vector database (FAISS/ChromaDB) to enhance park discovery beyond proximity, enabling intent-based results (e.g., shaded, less crowded, dog-friendly).

Designed an agentic workflow (LangChain / LangGraph) orchestrating geo queries, retrieval, ranking, and response generation with tool-calling and structured outputs.

Developed scalable Node.js/Express APIs for location storage and nearby park retrieval, integrating AI pipelines with clean service boundaries and reusable middleware.

Optimized performance using multi-layer caching (TTL-based request/response caching), reducing redundant API calls and improving response latency for frequent location queries.

Enhanced recommendation quality using hybrid search (vector + keyword), metadata filtering, and reranking techniques to improve relevance and user experience.

● Built a unified listing API reused across multiple features, enabling consistent filtering/sorting and reducing duplicate backend logic.

● Increased multi-API retrieval speed by 40% through optimized joins, indexing, Redis caching, and pre-aggregated SQL pipelines.

● Reduced frontend bundle size by 35% using memoization, code-splitting, and lazy loading to improve interaction speed.

● Increased dashboard responsiveness by 40% through virtualization and optimized rendering patterns.

● Reduced API load by 30% with pagination, caching, and debounce/throttle strategies.

● Secured the application with JWT authentication, refresh-token rotation, and role guards supporting 6 user roles.

● Designed scalable database schemas for multi-entity workflows and high-volume storage.

● Delivered 30+ mobile UI components using React Native and Expo for advanced media workflows.

● Built advanced media features including TTS synchronized highlighting and FFmpeg-based audio/video processing pipelines.

● Reduced infrastructure costs by Dockerizing backend services and deploying on AWS (EC2/S3/RDS) supporting 10K+ videos.

● Versioned models and exposed predictions, confidence scores, and evaluation metrics in an admin dashboard.

● Implemented secure authentication using JWT, RBAC, and 3 custom security filters to protect REST APIs.

● Built layered backend architecture (DTO/service/repository) with exception handling for maintainability.

● Reduced API response times by ~30% through SQL tuning, pagination, indexing, and caching.

● Built 25+ Angular components and modules to improve UI consistency and modularity.

● Delivered stable AWS deployments via EC2/RDS setup and secure VPC configuration.

● Improved UI consistency and maintainability by developing reusable frontend components

● Enhanced code quality by implementing backend logic in Java with JUnit testing and resolving key UI defects for smooth releases.